Background
Phase III double-blind RCT, N=413, patients with advanced HCC previously treated with sorafenib, Child-Pugh A, ECOG PS 0–1. Enrolled at 119 sites in 27 countries. Randomized 2:1 (pembrolizumab:placebo). Dual primary endpoints OS (prespecified final threshold P≤.0174) and PFS (threshold P≤.002 at interim). Stratified by region, AFP, ECOG PS, and macrovascular invasion/EHS.
Interventions and follow up
Arm A: Pembrolizumab 200mg IV q3wk + best supportive care (BSC)
Arm B: Placebo IV q3wk + BSC
Primary endpoint: OS and PFS (co-primary)
Median follow up: 13.8mo (pembrolizumab)
Arm B: Placebo IV q3wk + BSC
Primary endpoint: OS and PFS (co-primary)
Median follow up: 13.8mo (pembrolizumab)
Results
OS: 13.9 vs 10.6mo, HR 0.781 (95% CI 0.611–0.998), P=.0238 — prespecified threshold P≤.0174 not met
PFS (final): 3.0 vs 2.8mo, HR 0.718 (95% CI 0.570–0.904), P=.0022 — prespecified threshold P≤.002 not met
PFS (final): 3.0 vs 2.8mo, HR 0.718 (95% CI 0.570–0.904), P=.0022 — prespecified threshold P≤.002 not met
Adverse events
Overall: Grade ≥3 AEs 52.7% vs 46.3% (pembrolizumab vs placebo); treatment-related grade ≥3 18.6% vs 7.5%.
Hepatic/viral: No HCV or HBV flares identified; no new safety signals.
Hepatic/viral: No HCV or HBV flares identified; no new safety signals.
Conclusions
Pembrolizumab did not meet prespecified statistical significance for either OS or PFS in previously treated HCC. However, numerically meaningful trends were observed (OS HR 0.781, P=.0238) consistent with the phase II KEYNOTE-224 data, suggesting biological activity but insufficient power or enrichment to achieve regulatory approval in the 2L setting.
Key Limitations
Hierarchical thresholds were more conservative than conventional P<.05; both endpoints narrowly missed these thresholds, raising the question of whether the trial was adequately powered. The comparator was placebo — no standard second-line existed at the time, so the design was appropriate but may have underestimated the marginal benefit over modern supportive care. PD-L1 expression was not used for patient selection. Both OS (HR 0.781) and PFS (HR 0.718) showed numerically meaningful trends.
Clinical Context
Despite being a formal "negative" trial, KEYNOTE-240 followed the FDA's accelerated approval of pembrolizumab for 2L+ HCC based on KEYNOTE-224 (phase II, ORR 17%). The confirmatory phase III technically failed, yet pembrolizumab remains in use in the 2L setting. The landscape has since shifted with cabozantinib (CELESTIAL), ramucirumab (REACH-2 in AFP-high), and regorafenib (RESORCE) as established 2L standards, and nivo+ipi (CheckMate 9DW) entering 1L.
References